1. Field of the Invention
The present invention relates to a technology for determining a specific one type of statuses of an apparatus, such as an image forming apparatus, based on status data pertaining to the apparatus.
2. Description of the Related Art
When a failure that can be solved only by replacing parts or performing cleaning occurs in an apparatus such as an image forming apparatus, the apparatus inconveniently remains to be out of service until the failure is solved. Sudden failure of an electrophotographic image forming apparatus is likely to occur if regular maintenance is not appropriately performed. This is because an electrophotographic image forming apparatus contains a huge number of components has a complicated structure.
Not only wear resulting from normal operation but also other causes gradually degrade performance of an electrophotographic image forming apparatus and eventually lead to failure. The other causes are, for example, entrapment of a detrimental matter, such as paper dust, in the apparatus from outside, excessive agitation of toner, wear of a cleaning member such as a cleaning blade, degraded performance of a charging unit due to deposits on the charging unit, and random failure. The excessive agitation of toner, which can occur when the apparatus is operated in an unexpected manner or the like, results in an undesirable increase in viscosity of toner and/or separation of an external additive from toner. Such a failure of the apparatus manifests itself in the form of degradation in image quality in some cases. Examples of the degradation in image quality include streaks on an image extending in a moving direction of an image carrier, blurring of an image, streaks on an image extending perpendicular to the moving direction of the image carrier, spots on an image, a void, and dirty printing. Because the failure that can result in such degradation does not stop or hinder image forming operation performed by the image forming apparatus, the image forming apparatus continues to operate even in the anomalous status until a user of the apparatus visually checks a formed image and recognizes the anomalous status. The user who has recognized the anomalous status is required not only to repair the apparatus but also to perform image forming again to obtain properly-formed images. This is great waste of time and resources.
To this end, various methods of predicting occurrence of apparatus failure in advance have been proposed. Methods of predicting occurrence of a failure in an apparatus are generally classified broadly into a first group methods and a second group methods.
The first group methods include methods of monitoring information relating to an operation of an apparatus to make prediction depending on whether a period of service of the apparatus is near an end of an average usable period of the apparatus having been set in advance and the like. This approach is based on a premise that the longer an apparatus operates, the more likely a failure occurs in the apparatus as in the case where the apparatus is worn. More specifically, for example, an accumulative period of service of at least one of various components and devices of a photosensitive element and/or a developing device of the image forming apparatus is counted. A prediction that a failure is likely to occur in the image forming apparatus is made when the count value has reached a usable-limit count value. The usable-limit count value has been set in advance based on a result of a durability test or the like. Meanwhile, a usable period of an apparatus largely depends on its environment and operation method. However, because individual differences in environment and operation method are not taken into consideration in the first group methods, approaches of the first group are less accurate than those of the second group methods.
The second group methods include methods of making prediction based on pattern recognition by using condition data set as a pattern. The condition data set is specifically detected in an apparatus that is in a prefailure status. This method based on the pattern recognition can be performed, for example, by using multivariate analysis such as Mahalanobis-Taguchi system (MTS). In the second group methods, a prediction is made based on whether an apparatus is in a status that would lead to a failure. Accordingly, if it is possible to detect a specific prefailure status, this approach of the second group methods permits accurate prediction of a time when the apparatus actually requires a maintenance work without being affected by an environment and operation method of the apparatus.
Japanese Patent Application Laid-open No. 2005-017874 discloses a technique for predicting failure occurrence of an image forming apparatus by using the MTS. In this technique, a plurality of types of status data pertaining to an image forming apparatus is collected, an index value for use in determination as to whether a specific one type of the status of the image forming apparatus is anomalous is calculated from the status data, and whether a failure related to the specific type of the status is going to occur is predicted based on the index value. More specifically, in this technique, a large number of status data sets are collected from an image forming apparatus in a normal status or a test apparatus that has the same configuration as the image forming apparatus and in a normal status. A normal data set group (normal index data) is structured from the collected status data sets. A distance (index value) between the status data and an origin (average point of normal status data) is calculated. The distance indicates a relative position of the status data in a multidimensional space of the normal data set group. When the status data is deviated from a normal status to fall into a status where a failure is likely to occur, the correlation between the status data and the normal data set group in the multidimensional space is disturbed. In this case, the distance, i.e., the index value, increases. In contrast, when the image forming apparatus is in a normal status, the distance (index value) decreases. Hence, it is possible to determine how normal the image forming apparatus is based on the index value. In this manner, the technique disclosed in Japanese Patent Application Laid-open No. 2005-017874 permits to predict occurrence of a failure of an image forming apparatus in advance by detecting a relatively minor anomalous change that occurs antecedent to the failure. If occurrence a failure is predicted, a user can order a replacement part in advance or, in a case where repair is difficult to be performed by the user, order on-site repair by a service person in advance. This advantageously reduces downtime of the image forming apparatus. Furthermore, because only a component of which usable life is near its end and which has caused a minor anomaly is to be replaced, it is possible to avoid unnecessary replacement of another component whose remaining usable life is long, thereby avoiding an increase in maintenance cost.
Japanese Patent Application Laid-open No. 2005-227518 discloses a data collecting technique for collecting a plurality of data sets about a plurality types of statuses of an image forming apparatus. In this data collecting technique, the data sets are collected from a plurality of image forming apparatuses of the same model as that of a target image forming apparatus. The data sets are collected from the image forming apparatuses during their operation tests that are performed before shipping. A reference data group that includes all the data sets collected from the image forming apparatuses during the operation tests serves as an initial reference data group. An index-value calculating equation from which an index value for a specific one type of status is to be calculated is determined based on the initial reference data group. After an image forming apparatus delivered to a user has started operation, data is collected from the image forming apparatus and added to the reference data group at predetermined updating timing.
Some known methods determine a status of an apparatus by using a boosting method. The boosting method generally combines a plurality of less-accurate, weak discriminators into a highly-accurate discriminator. To predict various failures of an image forming apparatus by using this approach, whether a status of the image forming apparatus is a prefailure status is determined on a failure-by-failure basis. Examples of the failures include error operation of the image forming apparatus and defective printing. The prefailure status is a status that can serve as a sign of a failure of the image forming apparatus. To implement this approach, a plurality of types of status data is collected. Examples of the status data include a detection result of a sensor and data on operation control of devices that is converted into a numeric form. Whether the status data indicates a normal status or an anomalous status is determined by a weak discriminator on a status-data-by-status-data basis. A weight is assigned to each of determination results of the weak discriminators such that the more the status data is correlated with a failure, the greater the weight assigned to the status data is. By performing majority vote or the like of the weighted determination results, whether the image forming apparatus is in the prefailure status (anomalous status) or not (normal status) is determined.
Japanese Patent Application Laid-open No. 2002-262010 discloses a technique of displaying, when an anomaly is detected in an image forming apparatus, an indication of this effect on an operation panel.
When it is possible to predict occurrence of a failure in an apparatus based on determination as to whether the apparatus is in an anomalous status, a maintenance work can be performed before a failure actually occurs in the apparatus, thereby preventing a failure.
However, even when occurrence of a failure is predicted based on detection of an anomalous status, it is difficult to ensure that the failure is going to occur in near future without fail. Meanwhile, if unnecessary maintenance work is performed, it adds to the maintenance costs. Accordingly, even when a message is displayed to inform a user that an anomalous status has been detected as in the technique disclosed in Japanese Patent Application Laid-open No. 2002-262010, it is in many cases difficult to expect the user to perform a maintenance work before a failure actually occurs. However, if any trouble manifests itself to a user when a prefailure status is detected, the user can comprehend that the failure will occur in near future. This increases the possibility that the user performs a maintenance work. However, if nothing manifests itself to a user even when the prefailure status is detected, the user cannot comprehend that a failure will occur in near future. In this case, it is particularly difficult to expect the user to perform a maintenance work at a certain amount of cost before a failure actually occurs.